Machine learning methods for bipedal dinosaur tracks: Integrating different ichnological data - Code, data and models

Code, trained models and data to classify theropod and ornithopod tracks This work presents models for two machine learning (ML) tasks. The first task is the automatic segmentation of track photographs to create binarized version of the images (where the footprint is white and background is black). The second task is the classification of the tracks tasks, where the classifiers are divided into two classification levels: ichnogenus (Anomoepus, Caririchnium, Dinehichnus, Grallator, Iguanodontipus, Kalohipus, Kayentapus, Megalosauripus) and theropod/ornithopod. For the classification tasks we use two different types of ML models. The first group of models corresponds to deep learning models, using Convolutional Neural Networks (CNNs) to predict the most probable classification for the images. The second group of models uses morphometric data retrieved from the tracks to predict the most probable classification. CNN models are trained in three different data types: photographs, their binarized versions, and depth maps. All ML models are available in an user-friendly App in Hugging Face: Segmentation - https://huggingface.co/spaces/ML4Paleontology/Dinosaur-Track-Binarization-Assistant Classification - https://huggingface.co/spaces/ML4Paleontology/Dinosaur-Track-Classifier Data folder The data folder contains example images used to train and validate the CNNs and segmentation models, and tables with morphometric data from the tracks. All the datasets are separated by classification task and data type. Models and Code folder The models and code folder contains the trained models (checkpoints) and the code that was used to train the models. All the models are separated by classification task and data type. Acknowledgments The authors want to acknowledge the following institutions that have allowed the access to the analyzed specimens: Jurassica Museum, Museo Aragonés de Paleontologia, Museo de Ciencias Naturales de la Universidad de Zaragoza, Museo Numantino de Soria, and Sociedade de História Natural de Torres Vedras.